Opens in a new tabSkip to content
Agent LighthouseAgent Lighthouse

    Searches the text of every published page. The evidence sources themselves are not in this index — search all of them on the trusted sources page.

    GitHub ↗
    Browse checks and page contents
    answer-readiness/comparison-tables

    Comparison tables present

    What it checks

    Visible headings or captions containing compare, compares, compared, comparing, comparison, versus or vs establish comparison intent on every page type. An ordinary table, price or product classification does not establish that intent. Relevant pages need a visible table with a header row of at least two header cells and a data row of at least two data cells. Presentation tables do not count. All relevant pages covered: pass; some: warn; none: fail; no relevant pages: NA.

    Why it matters

    The check remains C/informative with weight 0. Existing evidence supports treating semantic tables as an optional structure convention, not as a proven AI citation advantage. No new consumer or scoring claim is introduced.

    Evidence

    Comparison tables

    Mechanism: Expressing comparative/multi-attribute data as a real HTML table (rather than styled divs or prose) increases the probability that an answer engine extracts the comparison and cites the page for comparative queries.

    Tables are included in the structural bundles that measured positive. GEO-SFE’s meso-structure — lists and tables at 25–35% of content — contributed 39.7% of a 17.3% citation gain (p<0.001, −3.1pp on ablation), and Zyppy’s ‘AI-ready Structure’ factor (explicitly naming tables) scores 8.6. The mechanistic argument is decent — TableRAG-style pipelines linearize tables to Markdown, so a genuine table survives extraction as aligned rows while a div-grid comparison degrades into unlabeled text fragments. Google does require that ‘important content is available in textual form’, which a real table satisfies and an image-of-a-table does not.

    Counter-evidence: The strongest counter-evidence in this domain. Format as a Prior measured a cross-model preference hierarchy in which ‘semantically rich formats such as texts and KGs are consistently favored over visually structured ones like infoboxes and tables’. Against plain text, tables scored 0.398, KGs 0.336 and infoboxes 0.235. Models systematically preferred prose when the same fact was available both ways. No vendor documents tables as a citation signal, and Google states there are no additional requirements or special optimizations for AI features.

    No study located isolates the causal effect of converting a styled div-grid into semantic table markup — the specific edit this audit would recommend. The widely circulated claim that ‘pages with well-structured comparison tables have a structural advantage’ traces only to SEO vendor blogs with no disclosed methodology. Keep the audit as an accessibility/extractability hygiene check (real table markup beats an image or a div grid), not as a citation-rate lever.

    Limits

    Intent detection covers explicit English headings and captions only. Prose comparisons without those labels remain NA. Row-oriented comparisons and other accessible table layouts are not recognized by this conservative shape check. A table elsewhere on a comparison page may satisfy the page-level check; association with a specific comparison section remains deferred.

    Sources